JP3691784B2 - Low frequency induction type high frequency plasma reactor - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates generally to wafer processing systems, and more particularly to a plasma reactor for a wafer processing process in which plasma is generated primarily by inductively coupled power.
[0002]
In the drawings, the first numeral of the reference number indicates the first drawing in which the component (member) indicated by the reference number appears.
[0003]
[Prior art]
Plasma etching and deposition are anisotropic, chemically selective, and can be processed under conditions away from thermodynamic equilibrium, making plasma etching and deposition in circuit fabrication attractive. It is a simple method. The anisotropic process allows the fabrication of integrated circuit patterns having sidewalls extending substantially perpendicularly from the edge of the masking layer. This is important in current and future ULSI devices where the etch depth, pattern width and pattern spacing are all equivalent.
[0004]
FIG. 1 shows a typical plasma processing apparatus 10 for wafer processing. This reaction apparatus is provided with an insulating-coated metal wall 11 surrounding the plasma reaction chamber 12. The wall 11 is grounded and functions as one side of the plasma electrode. A gas is supplied from the gas supply source 13 to the reaction chamber 12, and the gas is exhausted by an exhaust system 14 that forcibly exhausts the gas from the reactor in order to maintain a low pressure state suitable for the plasma process.
[0005]
The high frequency power supply 15 connected to the second electrode 16 electrostatically supplies power to the plasma in the reaction chamber 12. For processing, the wafer 17 is placed on or near the electrode 16. The wafer 17 is loaded into the reaction chamber 12 through a port such as a slit valve 18 and unloaded from the reaction chamber 12.
[0006]
In the plasma reaction apparatus, a 13.56 MHz high frequency power supply (RF power supply) is widely used. This is because this frequency is an ISM reference frequency (ISM means industrial, scientific, and medical fields), and the government-regulated emission limit of the ISM reference frequency is a non-ISM frequency, especially the communication band. This is because the regulation is more lenient than in the case of the above frequency. Because of this ISM standard, there is a lot of equipment used at that frequency, which further encourages worldwide use of 13.56 MHz. Other ISM reference frequencies are 27.12 MHz and 40.68 MHz, which are the first and second harmonics of the 13.56 MHz ISM reference frequency.
[0007]
The plasma is composed of two qualitatively different parts, that is, a quasi-neutral and equipotential conductive plasma body 19 and a boundary layer 110 called a plasma sheath. The plasma body consists of negatively charged and positively charged molecules of almost the same concentration as well as radicals and stable neutral molecules. The high frequency power supplied to the reaction chamber supplies energy to free electrons. Sufficient energy is transmitted to many of these free electrons, and as a result, ions collide with gas molecules to generate ions. The plasma sheath is a low conductivity region where the gradient of the spatial potential (ie, electric field strength) is large and electrons are insufficient. Such a plasma sheath is generated between a plasma body and an interface such as a wall of a plasma reaction chamber or an electrode.
[0008]
When the electrode is electrostatically coupled to a high frequency power source, the DC component V on the negative side of the voltage at this electrode _dc (Ie, DC bias) occurs (see, eg, H. Butler and G. S. King Fluid Physics, Vol. 6, page 1348 (1963)). This bias is the result of unbalanced electron and ion mobility and non-uniform sheath capacitance (capacitance) at the electrodes and walls. The magnitude of the sheath capacitance is a function of the plasma concentration, as is the shape of the plasma chamber and the relative area between the electrode and the wall in the chamber. Sheath voltages on the order of several hundred volts are generally made at the electrodes (see, for example, J. Coburn and E. Kay, substrate positive ion bombardment in RF diode glow discharge sputtering, Applied Physics, Vol. 43, 4965 (1972). See
[0009]
The DC component of the sheath potential at the powered electrode is beneficial for accelerating ions to a higher energy state in a direction substantially perpendicular to the electrode. Therefore, in the plasma etching process, the wafer 17 to be etched is placed on or slightly above the electrode so that a bundle of cations is projected substantially perpendicular to the wafer surface. This enables almost vertical etching of the non-protected area of the wafer. Several processes (silica (SiO 2) are required to produce the etching rate required for commercial etching processes (hereinafter “etch rate”). ₂ Such a high sheath voltage (and a high discharge voltage) is indispensable in the etching and the like).
[0010]
[Problems to be solved by the invention]
Transistor speed specifications and high integration in modern MOS integrated circuits require the use of shallow junctions and thin (on the order of 10 nanometers) gate oxide under a polysilicon gate that is thousands of angstroms thick. Unfortunately, such an IC structure is sensitive to bombardment by high energy (energy above 100 eV) ions as in the conventional plasma etching apparatus of FIG. It becomes difficult to avoid damage to the gate oxide during the etching process. It may be advantageous to operate at lower discharge power levels and discharge voltages since wafer damage is reduced as the sheath voltage associated with ion energy decreases. However, at 13.56 MHz capacitively coupled power, the voltage drop results in a proportional decrease in etch rate in many processes, thereby greatly reducing process efficiency.
[0011]
The etch rate in the silica and some silicon etch processes is a function of the ion bombardment power density transferred from the plasma to the wafer. Since this power is equal to the product of the electrode sheath voltage and the wafer ion current density, the wafer ion current density must be increased to maintain a substantially constant etch rate at the reduced sheath voltage. This requires increasing the plasma ion density near the wafer. Unfortunately, in a conventional plasma etching apparatus, the sheath voltage of the electrode and the ion density near the electrode are proportional to each other, and they are a monotonically increasing function of the amplitude of the high frequency voltage applied to the electrode.
[0012]
Thus, if the sheath voltage is reduced by lowering the voltage of the high frequency signal, the current density of the ion beam at the wafer will also be reduced, thereby increasing the etch rate more than at the sheath voltage or ion current. Cause a reduction in the rate of Therefore, in order for a soft etch process with a commercially sufficient etch rate (an etch process with a low sheath voltage on the wafer) to be performed, the wafer sheath voltage and ion density can be adjusted independently. Would be advantageous.
[0013]
One way to increase the etch rate by increasing the plasma ion density near the wafer is to use a magnet to create a magnetic confinement field that traps electrons near the wafer, thereby increasing the ion production rate on the wafer. To increase the associated density. The magnetic confinement field confines active electrons by advancing the active electrons spirally along a spiral orbit around the magnetic field lines.
[0014]
Unfortunately, the uniformity of the etch rate at the wafer surface is reduced, for example, due to the non-uniformity of the magnetic confinement field in a “magnetically enhanced” plasma etching system. E (electric field) x B (magnetic field) drift due to an electric field in and near the sheath also reduces etch rate uniformity in such systems. In order to improve the uniformity on the wafer surface in such a system, the wafer is rotated about an axis perpendicular to and centered on the electrode surface. This creates a cylindrically symmetric time-averaged field on the wafer with improved average uniformity on the wafer, thereby improving etch uniformity. However, such rotation creates undesirable mechanical movements in the plasma chamber that generate particulates and increase contamination.
[0015]
Another technique that can produce an acceptable etch rate at low ion bombardment energy is the recently developed electron cyclotron resonance plasma generation method. This technique has applications for wafer cleaning, etching, and deposition processes. In this technique, plasma is generated using a microwave power source and a magnetic confinement structure. Unfortunately, when applied to etching or chemical vapor deposition, this method produces fine particles at high levels, low etch rate uniformity in the radial direction, and low efficiency.
[0016]
Since the fraction of energy devoted to radical generation increases rapidly above about 0.13 Pascal, the pressure of the system must be kept below that level. This is (1): a very high pumping speed (more than 3,000 liters per second, which is 10 times the volume of a normal type) and the extremely low pressure required for this process Requires an expensive device comprising: a vacuum pump system that produces (0.013-0.13 Pascal); and (2) a huge magnetic confinement system that sometimes includes large electromagnets.
[0017]
Yet another technique for increasing ion density is to use a microwave plasma generator that generates ions in a region of at least 10 centimeters on the wafer. These ions flow into the space on the wafer and contribute to the ion density of the wafer. However, this method tends to generate a large amount of free radicals and only generates a few milliamperes of ion current density per square centimeter on the wafer.
[0018]
Joseph Frezinger and Horst W. Loeb's "Neutral Molecule Injector for Fusion Reactor RIG" Nuclear Energy and Nuclear Technology, Vol. 44 (1984) No. Pages 1,81-86 provide an additional amount of power required to generate a neutral beam of particles and provide a balance point in the energy generation of the tokamak fusion reactor. This beam is created by generating an ion beam with inductively coupled power and neutralizing the beam by passing a gas before entering the fusion reactor. The ion beam is extracted by a direct current field instead of the high frequency field in this application.
[0019]
J. et al. In the reactor shown in Fresinger et al.'S paper entitled "High-Frequency Ion Power Supply for Reactive Gas Material Processes (9th International Conference on Gas Discharge and its Applications, 19-19-23 1988)" In order to heat the electrons, power is supplied into the reaction chamber, and the ion beam to the wafer is generated by a direct current field instead of a high frequency field.
[0020]
[Means for Solving the Problems]
Based on the preferred embodiment described, a plasma reactor is shown. In the plasma reactor, a low frequency (0.1 to 6 MHz) high frequency power supply (RF power supply) is inductively coupled to the plasma to supply ionization energy of the gas in the vicinity of the electrode holding the wafer, Moreover, a lower power high frequency voltage is applied to the electrode to control the ion bombardment energy of the wafer on the electrode. The wafer is placed on or directly above the electrode surface for processing.
[0021]
The plasma reactor comprises a non-conductive reaction chamber wall surrounded by an induction coil connected to a low frequency RF power source. The split Faraday shield is disposed between the induction coil and the side wall of the reactor, and surrounds the reactor and substantially eliminates the generation of displacement current between the induction coil and the plasma reactor. In fact, this shield significantly reduces the electrical coupling of low-frequency RF electric fields to the plasma (the paper entitled “Glow Discharge Phenomena in Plasma Etching and Plasma Deposition” by JL Bossen) Technology 126, No. 2, February 1979, page 319). As a result, the ion bombardment energy of the reactor wall and the associated etching and sputtering of the reactor wall are substantially eliminated, and the modulation of the wafer sheath voltage at low frequencies is reduced.
[0022]
The Faraday shield is movable so that the capacitance (capacitance) between the plasma and the shield can be changed. This Faraday shield is placed almost in contact with the outer wall of the reaction chamber, creating a high capacitance during the wafer processing process. This reduces the high frequency plasma potential, thereby reducing the plasma etching of the reactor walls. The increased spacing between the Faraday shield and the reaction chamber wall, which produces reduced capacitance, is also available outside the wafer etch to produce increased radio frequency and time average plasma potential levels, thereby A higher ion bombardment energy is generated that allows the reactor wall to be cleaned with controlled etching levels.
[0023]
Preferably, the Faraday shield is moved radially to change the capacitance, but the capacitance can also be changed by moving the Faraday shield in the vertical direction. In an embodiment in which the shield is movable in the vertical direction, the shield should not be allowed to move in the vertical direction so that it no longer exists between the reaction chamber and each induction coil. A conductive sheet may be included at the top of the reaction chamber to increase the plasma capacitance to the effective high frequency ground electrode provided by the reactor wall. This plate may also be movable to change the capacitance between the plasma body and this part of the Faraday shield.
[0024]
A direct magnetic field may be included to promote ion generation at low pressures by confining electrons away from the reaction chamber walls. Under low pressure, the electrons have an increased mean free path that increases the loss ratio from the reaction chamber by collision with the reaction chamber wall. This magnetic field causes electrons to enter a spiral path that increases the rate of ionization collisions in the reaction chamber before collision with the wall.
[0025]
In order to bounce electrons back into the plasma, a distributed magnetic field that is stronger near the top of the reaction chamber may be included, thereby preventing loss of electrons at the top of the reaction chamber wall. This magnetic field (approximately tens of thousands of Tesla near the upper part of the reaction chamber) is arranged by an arrangement of permanent magnets arranged in the upper part of the reaction chamber and having an alternate magnetic field direction, or by a solenoid coil in which a direct current flows, Alternatively, it is generated by a ferromagnetic disk.
[0026]
The inductively coupled high frequency power is supplied to a level of 10 kW at a frequency in the range of 0.1 to 6 MHz, depending on the size of the reaction chamber. The voltage applied to the electrode is at a frequency higher than the reciprocal of the average time that ions traverse the sheath of the electrode. The frequency f of this voltage signal _h Preferred are all ISM standard frequencies, ie 13.56 MHz, 27.12 MHz, 40.68 MHz. Higher frequencies will require higher frequencies to produce ion bombardment energy that is less widely dispersed.
[0027]
The sheath of the electrode has a strong electric field that is approximately perpendicular to the wafer surface, thereby producing a substantially perpendicular ion bombardment and a substantially perpendicular or controlled tapered wafer etch. The amount of capacitively coupled power provided to the electrodes is much less than the power inductively supplied to the plasma. Therefore, the average ion current in the wafer is primarily determined by the inductive coupling power. Then, due to the Faraday shield, the average ion energy in the wafer becomes approximately a function of only the amplitude of the high frequency signal (rf signal) to the electrode.
[0028]
In contrast, in the typical plasma reactor shown in FIG. 1, both the average ion density (generally somewhat lower) and energy are controlled by the amplitude of the radio frequency signal to the electrodes. Therefore, the inductively coupled reactor makes it possible to reduce the sheath voltage and increase the ion density. Also, the sheath voltage and ion density can be changed separately. As a result, a soft etch with a commercially acceptable etch rate is achieved, which damages recent types of integrated circuits that can be damaged by ions with impact energy on the order of 100 eV or more. I will not let you.
[0029]
The electromagnetic field in the inductively coupled plasma reactor produces a very uniform plasma ion density distribution on the wafer, realizing a very uniform wafer processing process. The inductively generated electric field is approximately cylindrical and therefore accelerates the electrons approximately parallel to the reactor sidewall. Due to the conductivity of the plasma, the electric field strength decreases rapidly away from the sidewall, and electron acceleration occurs mainly in the region near the sidewall.
[0030]
As the electrons increase in velocity, their inertia describes a trajectory that includes a series of elastic collisions with the molecules and / or a grazing contact with the sidewall sheath. Such a collision or contact repels electrons into the plasma body. This results in significant electron acceleration only in the vicinity of the wall, but also generates ions everywhere in the reaction chamber. These electron and ion diffusions, as well as the radial E × B drift of the electrons, produce a radially symmetric ion density with a very uniform density in the vicinity of the wafer. The reaction chamber is kept at a low pressure (generally on the order of 0.13 to 3.9 Pascals) in order to promote scattering of electrons away from the region where the electrons gain energy near the side walls.
[0031]
This design is also very effective in coupling power to ion generation and therefore offers significant advantages over other reactors for wafer processing processes performed by ions in the plasma. (For example, a paper entitled “RF ion power supply RIM10 for material processing with reactive gas” by J / Fresinger et al., 9th International Conference on Gas Discharge and its Applications, September 19-23, 1988) reference). The importance of this is as described below.
[0032]
The high frequency power to the plasma creates neutral radicals, ions, free electrons, and excited states of molecules and atoms by free electrons. Vertical etching with reactive ions is advantageous for reaction chambers that divert much of the RF power to ion generation. Due to excessive radical concentration, reactions at the wafer surface by radicals are the target fabrication process. Reducing the relative production of free radicals by the plasma is beneficial in many applications because it can be harmful to the plasma. Therefore, this plasma reactor is particularly suitable for reactive ion etching processes or other processes that are favored by high ion concentrations or are degraded by significant free radical concentrations.
[0033]
This reactor also requires much less capacitively coupled power than conventional plasma reactors. This system uses high frequency power on the order of several hundred watts, compared to 500-1000 watts for a conventional plasma reactor where all power is electrostatically coupled. The system also has the ability to control ion flow and ion collision energy separately.
[0034]
In the conventional plasma reactor shown in FIG. 1, the amplitude of the high-frequency signal applied to the electrode 16 controls not only the ion density in the plasma but also the sheath voltage of the electrode. To achieve soft etching (ie, wafer ion bombardment energy of about 100 volts or less), electrostatically applied high frequency power is greater than that traditionally used in such reactors. Should also be lowered.
[0035]
Unfortunately, this reduction in electrostatic applied power not only reduces the voltage drop across the sheath, but also reduces the ion density at the sheath. Even under high RF voltages on the electrodes, such capacitively coupled power results in only a relatively low ion density. Since the wafer etch rate is proportional to the product of the ion density in the sheath and the voltage drop across the sheath, the wafer etch rate decreases faster than either of these two parameters. Thus, soft etching results in a reduction in efficiency that is incompatible with commercial integrated circuit fabrication processes.
[0036]
It can be shown with reference to FIGS. 2 and 3 that the sheath voltage in this system is constrained by the amplitude of the high frequency signal applied to the electrodes. The capacitor 21 between the high frequency power supply 15 and the electrode 16 enables this sheath voltage to have a DC component. This direct current component is produced by a synergistic effect of the uneven region of the electrode and the uneven mobility of electrons and ions. Each plasma sheath is electrically equivalent to a parallel combination of resistor, capacitor and diode. The electric field across the sheath is 10 ^Four Most electrons are repelled out of the sheath region that generates a large sheath resistance of the order of ohms.
[0037]
The electrostatic component of sheath impedance as a function of increasing frequency is small enough to make sense at about 500 kHz and can be ignored below that frequency. At frequencies above 500 kHz, the sheath resistance is very large and can be ignored. This is the case for the high frequency sheath voltage component at the frequency of the electrostatically coupled power.
[0038]
In the equivalent circuit of FIG. 2, the effects of electron mobility much greater than the plasma and ions in the sheath are modeled by diodes 24 and 28. Thus, if the plasma is negative with respect to all electrodes in close proximity to the plasma, the electrons in the plasma will effectively short to that electrode. Therefore, the sheath impedance is modeled by elements 22-24 and 26-28. The plasma break is the high frequency f used for the RF voltage applied to the electrodes. _h It is modeled as a low impedance resistor 25 that can be neglected (preferably one of the ISM frequencies 13.56 MHz, 27.12 MHz or 40.68 MHz).
[0039]
FIG. 3 shows the frequency f applied to the electrode _h The relationship between the 220 volt peak-to-peak high frequency signal 31, the plasma resultant voltage 32, and the electrode sheath voltage 36 is shown. Sheath capacitance C _S1 And C _S2 Is the frequency f of the electrostatic coupling power _h Resistance R _S1 And R _S2 Can be ignored, and diodes 24 and 28 can be ignored except for a short interval in each period of signal 31. Therefore, under the most working conditions, the plasma equivalent circuit is converted to an electrostatic divider and the plasma potential V _P And capacitance C _S1 as well as _CS2 The high-frequency component of the voltage across the line is almost in phase and the magnitude is V _P = V _rf ・ C _S2 / (C _S1 + C _S2 ).
[0040]
In a typical reactor with a wall area several times the electrode area, the sheath capacitance C at the wall _S2 Is the sheath capacitance C at the electrode _S1 Is about 10 times. Therefore, for the 220 volt peak-to-peak high frequency signal 31, the plasma potential V _P Is about 20 volts peak-to-peak. Since the signals 31 and 32 are in phase, the peak 33 of the signal 32 is aligned with the peak 34 of the signal 31. Because of diode 24, the minimum voltage difference between signals 31 and 32 (which occurs at each peak 34) is kT. _e / E or so. Similarly, to prevent a short circuit of plasma to the reactor wall, V _P Is at least kT on the positive side of the ground 35 _e / E must be present.
[0041]
Due to these various conditions, the average sheath voltage 36 of the electrode (that is, the DC component of the high-frequency signal 31) is approximately −90 volts. The DC component of the sheath voltage is -V _rf ・ C _S1 / (C _S1 + C _S2 ) / 2, approximately equal to V _rf Is the peak-to-peak intensity of the high-frequency voltage. Since the electric field component of the high-frequency signal is substantially perpendicular to the electrode, the sheath voltage changes directly with the high-frequency signal intensity. This means that the direct current component 36 of the voltage 31 is directly related to the peak-to-peak amplitude of the high-frequency voltage applied to the electrode.
[0042]
The ion current density at the electrodes of a conventional plasma reactor is proportional to the ion density in the plasma, which decreases as the power decreases, and the current density also decreases when the amplitude of the high-frequency voltage is reduced to reduce the sheath voltage. Will do. Therefore, in the plasma reactor of FIG. 1, even when the voltage is lowered to perform softer etching, the current density in the wafer cannot be increased in order to maintain the etching power.
[0043]
The voltage drop across the sheath at the electrode is equal to the difference between the applied high frequency signal 31 and the plasma voltage 32. This voltage drop varies from 0 volts to about -220 volts. Ion is 1 / f of high-frequency signal _h Even if the sheath passes through the sheath at a time interval shorter than that of the period, if it passes through the sheath near the peak 34 of the high-frequency signal 31, the impact energy becomes almost zero. Such low energy bombardment ions do not necessarily draw a trajectory that is substantially perpendicular to the wafer surface and can therefore reduce the vertical etching of the intended wafer.
[0044]
Therefore, the period 1 / f _h However, it is important that no more than half of the average time an ion passes through this sheath. Since this transit time is about 1 / 500,000 second or less, f _h Must be at least 4 MHz. For higher ion density and low sheath voltage, the period 1 / f _h Becomes less than 0.1 microsecond (μs). Because of lenient restrictions on ISM frequency, f _h Is preferably equal to one of the ISM frequencies 13.56 MHz, 27.12 MHz, 40.68 MHz.
[0045]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 4 shows a plasma reactor in which the sheath voltage and the ion flow density at the wafer can be adjusted independently. This reactor also produces a very uniform distribution of ion flow density and voltage across the wafer, as well as the rate of ion production and free radical production in the plasma for purely electrostatic or higher frequency inductive discharges. It is possible to raise the ratio between. Therefore, this system is particularly beneficial for applications where the ratio of ions to free radicals is large.
[0046]
Above the base 40 is a cylindrical reaction chamber wall 41 surrounding a plasma reaction chamber 50 (see FIG. 5). The reaction chamber wall 41 is 7-30 centimeters high and has a side diameter that depends on the diameter of the wafer being processed. In a wafer processing system having a diameter of 15 cm, the reaction chamber wall has a side diameter of about 25 to 30 cm, and in a wafer processing system having a diameter of 20 cm, the reaction chamber is 30 to 30 cm. It has a side diameter of 38 cm. The reaction chamber wall 41 is made of a nonconductive material such as quartz or alumina.
[0047]
The surrounding wall 41 is an induction coil 42 itself connected to a first high-frequency power source 43 (hereinafter referred to as “RF power source”) via a traditional impedance matching circuit or transformer (transformer) 44. RF power supply by traditional matching circuit 44 using commercially convenient reactance values or by a transformer that matches the inductive impedance (typically 10 ohms or less) to the impedance of power supply 43 (typically 50 ohms) This coil is wound only a little (on the order of 2 to 8 turns) to produce an inductance that conveniently matches 43. The matching circuit is designed to substantially eliminate reflections of power returning to the power supply 43.
[0048]
This induction coil generates in the reaction chamber 50 an axisymmetric high frequency magnetic field whose axis is substantially vertical and a substantially cylindrical electric field. These two magnetic and electric fields are rotationally symmetric about the central axis A. This rotational symmetry contributes to the uniformity of wafer processing.
[0049]
Due to the high conductivity of the plasma, the inductively coupled field is the frequency f of the inductively coupled RF field. ₁ Is substantially limited to the region adjacent to the sidewall having a thickness δ (on the order of 1 centimeter) proportional to the square root of the electron density in the plasma divided (divided) by. In larger systems, to increase the thickness of this region that accelerates electrons, f ₁ Is lowered.
[0050]
Within this region, a cylindrical electric field accelerates electrons in the circumferential direction. However, due to this accelerated inertia of the electrons, the electrons will graze the electric field of the sheath at the sidewall. Such faint contact reflects many of the electrons from the wall. Some of the electrons strike the wall to produce secondary electrons. Elastic collisions with gas molecules cause electrons to diffuse throughout the reaction chamber. Since the inductively generated electric field is limited to the distance δ (resistance film thickness) from the side wall, the electron heating remains only in this region. To make the ion density across the wafer more uniform, the pressure is kept low (usually 0.13 to 3.9 Pascals), and electrons heated near the wall diffuse quickly from the wall and become nearly uniform. Ionization and resulting ion density at the wafer surface.
[0051]
Reactor radius R, frequency f ₁ , And inductively coupled power is selected to produce a cylindrical electric field with a peak-to-peak amplitude of 1-10 volts / cm. This results in a vibrating electronic path having an amplitude of 3 centimeters or more, and the mean free path of these electrons is about the amplitude of the electronic vibration or less. The power supply 43 supplies power at a frequency in the range of 0.1 to 6 MHz and an output up to 10 kW.
[0052]
The second RF power source 51 (shown in FIG. 5) is an electrode at one of the ISM (industrial, scientific, medical) standard frequencies (ie, 13.56 MHz, 27.12 MHz, 40.68 MHz). High frequency power is supplied to 52. In the example of FIG. 1, the high-frequency power source generates a direct current sheath voltage along with the electrode 52. Its power level ranges from less than 100 watts to hundreds of watts (up to 500 watts), and the impact of capacitively coupled radio frequency signals on ion density is much greater than that of inductively coupled power from the power supply 43. Very few. This power level is somewhat lower than the power level generally supplied to the electrodes of the plasma reactor. This power level is kept low to produce soft collisions of the wafer with ions (ie, kinetic energy of 100 ev or less). The low power level to this electrode also means that the ion density is mainly determined by the RF power source 43. This is advantageous when controlling decoupling of ion density and sheath voltage.
[0053]
The circumferential direction of the inductively generated electric field makes this electric field parallel to the electrode so that the path integral from the plasma body to the electrode along the normal to the electrode is zero. As a result, unlike the plasma reactor of FIG. 1, there is no high-frequency component of the sheath that generates an RF time-varying potential difference between the plasma body and the electrode. This substantially eliminates the coupling of the low frequency induced RF field to the electrode potential. Therefore, the sheath voltage of the electrode 52 is determined only by the RF power source 51.
[0054]
Surrounding the side walls of the reactor corresponds in this embodiment to a grounded Faraday shield 45 consisting of a dozen conductive plates 46 that become the side walls. Each Faraday shield conductive plate 46 is spaced from neighboring plates by a gap 48. These gaps are required to allow the induction high frequency magnetic field to penetrate through the reaction chamber 50. In order to prevent the generation of current in the circumferential direction in the Faraday shield, at least one gap is required. According to Lenz's law, such circumferential currents are strongly repelled by changes in the magnetic field in the reaction chamber 50, resulting in essentially countering the desired action of the coil 42 current in the reaction chamber 50. become.
[0055]
This Faraday shield also provides the same function as the grounded conductive wall of the reactor of FIG. That is, the Faraday shield allows the capacitively coupled RF field to be removed from the reaction chamber 50 so that the capacitively coupled RF field does not fall outside the reaction chamber and interfere with other devices or deviate from federal emission standards. Restrict to within. This shield also provides a return path for high frequency current from the electrodes created by the capacitively coupled power supply 51.
[0056]
When the Faraday shield 45 is adjacent to the reactor wall, the RF frequency f of the power source 43 ₁ Plasma potential at _P It is possible to greatly reduce the amount of time change in the. This is the ion density and average sheath voltage V _dc It is important to separate the influence of the first RF power source 43 and the second RF power source 51 on the. At the power level applied to the coils 42, the large inductance (on the order of 1-100 microhenries) of these coils results in a high voltage at one or both ends of the coils. If there is no Faraday shield, the high voltage end 47 of the coil 42 is electrostatically coupled to the plasma body and the frequency f of the power source 43 ₁ V at _P (See, for example, JL Bossen “Glow Discharge Phenomena in Plasma Etching and Plasma Deposition”, Journal of Electrochemical Society, Solid State Science and Technology, Vol. 126, No. 2,319.) ).
[0057]
The width of the gap 48 is narrower than the minimum distance between the conductive blade 46 and the coil 42 so that the coil 42 is not electrostatically coupled to the plasma body via these gaps (see the Bossen document). . If electrostatic coupling to such a plasma body is not hindered, this plasma potential V _P Changes in the RF will appear as changes in the sheath voltage (and hence ion energy) at the same frequency. Furthermore, this electric field will reduce the symmetry of the etching if it is hardly excluded by the Faraday shield.
[0058]
The Faraday shield 45 also has a sheath capacitance C of the plasma sheath adjacent to the wall 41 of the plasma reactor. _S2 The value of is greatly affected. If this Faraday shield is not present, effective grounding of the capacitively coupled radio frequency signal is provided by the environment surrounding the RF induction coil or its reaction chamber, and hence other existing in the vicinity of the reactor. Will be influenced by the object. Furthermore, these objects are generally at a distance large enough to be treated as having an infinite effective ground condition. This means that the C on the side wall and the top wall _S2 For C as in FIG. _S1 C, not 10 times _S1 1/10 or less. As a result, the plasma potential V _P And the high-frequency signal are much closer to the relationship shown in FIG. 7 than the relationship shown in FIG.
[0059]
In FIG. 7, assume again that the high frequency voltage (signal 71) has a peak-to-peak amplitude of 220 volts. C _S1 Is C _S2 The plasma voltage signal 72 has a peak-to-peak amplitude of 200 volts. Plasma voltage V _P The peak 73 of the high-frequency voltage signal 71 is again aligned with the peak 74 of the high-frequency voltage signal 71, and the interval between both peaks is again kT. _e Up to several times / e. Similarly, V _P The distance between the indentation and the ground is kT _e / E (usually a few volts). Therefore, the plasma voltage signal 72 has a DC component 76 on the order of 100 volts. This is because the plasma voltage signal 32 is about 10 volts kT. _e Contrast with FIG. 3, which has a DC component with an offset of about / e.
[0060]
This greatly increased DC component between the wall and the plasma body results in unacceptable levels of wall etching or sputtering by ions in the plasma. Such action not only damages the reaction chamber walls, but also depletes reactive gases and introduces contaminants into the plasma that interfere with the wafer fabrication process within the reaction chamber. However, if there is a Faraday shield 45 that is spaced somewhat away from the wall 41, the effective ground electrode capacitance increases and C _S2 Is C again _S1 Several times larger than the high frequency signal and plasma voltage V _P The relationship with is as shown in FIG. 3 instead of FIG.
[0061]
In fact, usually capacitance C _S2 The distance between the two plates (ie, plasma and conductive wall) is about 0.1 cm. In the reactor of FIG. 4, when the Faraday shield is placed near the wall 41, the capacitance C _S2 Is increased by the thickness of the wall 41 divided by the dielectric constant (greater than 4), which is equivalent to a vacuum gap of 0.075 centimeters. Therefore, the wall capacitance C _S2 Is slightly larger than half the capacitance in a reactor of the type shown in FIG. 1 having a size comparable to that of FIG.
[0062]
The conductive plate 46 can move in the radial direction by about 1 cm or more, and the capacitance C _S2 Is reduced by moving the conductive plate 46 away from the wall and over a range of approximately 0.1-10. _S1 / C _S2 Change the ratio. Since these conductive plates are moved near the wall 41 during the wafer processing process, the reaction chamber wall etching and contaminant by-products are minimized. Outside of the wafer processing process, the conductive plate is moved 1 cm or more away from the wall to periodically etch the wall and clean the wall. Residue debris from the reaction chamber cleaning process is removed from the reactor before further wafer processing processes are performed.
[0063]
FIGS. 5 and 6 are a side sectional view and a top view, respectively, showing prominent features of the reactor 40. Immediately outside the top of the reaction chamber 50 is a grounded conductive plate 53 that provides the top of the reaction chamber 50 with substantially the same function as the Faraday shield 45 has on the side of the reaction chamber.
[0064]
At the top of the reactor, there is a set of magnets 54 with the north poles alternately facing downward. The ferromagnetic reflector plate 55 helps to repel the magnetic flux of the magnetic field produced by the two outermost magnets. The magnet is preferably a permanent magnet, since this type of magnet provides a sufficient magnetic field economically. This arrangement creates an alternating magnetic field sequence of about 0.01 Tesla at the top of the reaction chamber 50, acting like a magnetic mirror that bounces electrons toward the plasma body.
[0065]
The magnetic field generated by these magnets enters the reaction chamber by a distance that is approximately twice the distance between these magnets (about 2 to 3 cm). In another embodiment, the linear arrangement of the magnets is replaced with a set of concentric annular magnets, and the adjacent north magnets are arranged with their north poles facing away from the vertical direction. May be. In yet another embodiment, a flat disk of ferromagnetic material with a north pole oriented vertically, or a single disk to create a magnetic mirror with a magnetic field on the order of tens of thousands of Tes near the top of the reaction chamber A ring DC solenoid may be used. Embodiments using magnetic disks are preferred because they are simple and inexpensive and retain the radial symmetry of the reactor. In contrast, the lack of radial symmetry by the magnet 54 of FIG. 5 may slightly reduce radial symmetry in wafer etching.
[0066]
A conductive coil 56 connected to a DC power source 57 is present outside the proximal end or upper part of the reactor wall 41 to generate an arbitrary DC magnetic field and contain electrons away from the sidewall. The magnetic field magnitude of this coil can be on the order of 0.0001-0.01 Tesla.
[0067]
The plasma reactor of FIGS. 4-6 exhibits a much improved effect compared to many other existing reactors. However, a plasma reactor using a microwave power source (described in the background of the invention) is only a few milliamps / cm. ² In contrast to a current density of 50-100 milliamps / cm. ² Up to. The test is SF ₆ , CF ₂ Cl ₂ , O ₂ In addition, the high current is generated in various reactive gases such as argon.
[0068]
This indicates that more power is devoted to the generation of ions instead of being directed to the generation of neutral fragments as in other plasma production processes at pressures of 1 milliTorr or higher. Such neutral fragments do not contribute to the current. This is important because only ions will impact the wafer in the vertical direction to form a nearly vertical wall. Having the ability to produce a very low sheath voltage in the wafer means that by lowering the sheath voltage to below 20-30 volts, the underlying 10 nanometer thick SiO 2 ₂ It means that a 400 nanometer thick polysilicon gate can be etched vertically without damaging or etching the gate insulator.
[0069]
The reactor 40 includes a gas supply source 49 and an exhaust port 58, which exhaust port 58 includes a pump for exhausting plasma processing products and maintaining a pressure at a predetermined level. Parts. In general, the pressure is kept on the order of 0.13 to 3.9 Pascals to promote electron diffusion from the electron heat field near the sidewalls into the bulk. Even at this pressure, the inductively coupled power is primarily directed to ion generation.
[0070]
In contrast, other plasma systems, such as microwave plasma systems, produce relatively more free radicals under pressures above about 0.13 Pascal. If the microwave plasma reactor mainly generates ions, the pressure needs to be about one hundredth of a pascal or less. This requires the reactor pump to have a speed much greater than a few Torr liters per second. This high pump speed requires the use of a cryogenic pump tightly attached to the reaction chamber or a turbo pump with a large port in the reaction chamber.
[0071]
In contrast, the reactors disclosed herein can operate at higher pressures and only require pump speeds on the order of tens of pascal-liters per second. This means that the space around the reaction chamber will not be disturbed, the wafer operation will not be disturbed, or a much smaller pump that will not get in the way around other reaction chambers will be in time. Such pumps also do not require regeneration and do not have the safety challenges as in cold pumps.
[Brief description of the drawings]
FIG. 1 shows the structure of a typical plasma reactor.
FIG. 2 is an equivalent circuit of a plasma reactor in which electric power is electrostatically coupled to a reaction chamber.
FIG. 3 shows a high frequency signal applied to an electrode and a plasma voltage V. _P And sheath voltage V _dc The relationship between is shown.
FIG. 4 is a side view of an inductively coupled reactor according to one embodiment of the present invention.
FIG. 5 is a side sectional view of the inductively coupled reaction device of FIG. 4;
6 is a top view of the reaction apparatus of FIG. 4. FIG.
FIG. 7C _S1 Is C _S2 Is greater than the plasma voltage V _P And the high-frequency voltage applied to the cathode.

Claims (11)

A plasma reactor for processing a semiconductor substrate,
A reaction wall (41) surrounding a reaction chamber (50) for generating plasma to produce at least one plasma product in the processing of a semiconductor substrate;
Means for connecting a gas supply source (49) and a gas exhaust system (59) to the reaction chamber;
A high frequency power source (43);
An induction coil (42) disposed adjacent to the reaction wall (41) outside the reaction chamber (50) and inductively supplying the high-frequency power to the reaction chamber;
A split Faraday shield (45) disposed between the induction coil and the reaction chamber, the split Faraday shield passing high frequency power from the induction coil through the split Faraday shield to sustain the plasma. A plurality of substantially non-conductive gaps (48) configured to inductively supply and prevent formation of repulsive currents repelling changes in the magnetic field in the reaction chamber (50), The width of the gap is the split Faraday shield that is narrower than the minimum distance between the split Faraday shield and the induction coil while the semiconductor substrate is being processed,
And a support for positioning the semiconductor substrate such that the semiconductor substrate is exposed to at least one plasma product during processing.   2. The plasma reactor according to claim 1, further comprising means for changing a level of an electrostatic shield supplied by the split Faraday shield.   The plasma reactor according to claim 1, further comprising means for changing the capacitance between the split Faraday shield and the plasma with respect to the capacitance between the semiconductor substrate and the plasma.   The plasma reactor according to claim 1, further comprising means for changing a capacitance between the split Faraday shield and the plasma. 5. The plasma reaction apparatus according to claim 1, wherein the split Faraday shield is composed of a plurality of conductive plates movable in a radial direction by a distance of 1 cm or more. 6. The plasma reaction apparatus according to claim 1, wherein the split Faraday shield is spaced from the reaction wall by less than 1 centimeter while the semiconductor substrate is being processed.   In a method for processing a semiconductor substrate in a plasma reactor, a step of supplying a gas to a reaction chamber (50) having a reaction wall (41) and a step of providing an induction coil (42) adjacent to the reaction wall And shielding the gas in the reaction chamber with a split Faraday shield (45) between the induction coil and the reaction wall, wherein the split Faraday shield has a plurality of substantially non-conductive gaps. And a step of positioning the split Faraday shield so that a width of the plurality of gaps is narrower than a minimum distance between the split Faraday shield and the induction coil while the semiconductor substrate is being processed. And in the gas through the split Faraday shield to maintain the plasma in the reaction chamber The method comprising the steps of inductively supplied electric power, and forming at least one plasma product in the processing of semiconductor substrates, and a step of exposing the at least one plasma product semiconductor substrate during processing.   8. The method of claim 7, further comprising changing a capacitance between the split Faraday shield and a plasma. The method of claim 7, further comprising a method the step of changing the capacitance between the split Faraday shield and the plasma relative to the capacitance between the semiconductor base plate and the plasma.   The method according to any one of claims 7 to 9, wherein the split Faraday shield is composed of a plurality of conductive plates, and further comprises a step of moving the plurality of conductive plates in a radial direction. The method according to any one of claims 7 to 10, wherein the split Faraday shield is arranged at a distance of less than 1 cm from the reaction wall while a semiconductor substrate is being processed.

Images